# Destructive VTMV Subgroups - A Destructive Variable Test Method to Validate

## Who is right?

• ### Original Poster

Votes: 0 0.0%
• ### Colleague

Votes: 0 0.0%

• Total voters
0
M

#### mwb0585

First time post here... and yes, I've searched the forums, but I'm still posting.

I have a destructive variable test method to validate. Specifically, its a cable tensile test. The minimum tensile force is 17lbs.

In order to do a destructive Gauge R&R, we must create homogenous subgroups. My idea was to take successive samples from the cable spool, inspect them, and if there are no imperfections, to just call them "identical." Do this several times to get several different subgroups. (Alternatively, we could also use different lot numbers to create the subgroups.)

My colleague is worried that if we do this there will not be enough part-to-part variation between the different subgroups, and that Minitab will not distinguish them as different groups.

He suggests we use a different cable, one with a minimum tensile force of 20lbs, to create some of the subgroups. This way there will be at least some distinguishable subgroups.

I say using a completely different cable will artificially lower the %P/T calculation, by making the tolerance much wider, since now the average of all the samples will be higher compared to the spec. (One sided tolerance is calculated as Sample Avg - LSL). (minitab_dotcom/support/documentation/Answers/One-Sided%20Tolerance.pdf) [sorry, i cant post links yet]

He says it doesn't matter, because really the tolerance should be 17 --> infinity (not 17 --> average of all samples). He insists that companies have validated cable tensile tests by using subgroups of different cable widths.

Finally, my questions: who is right?

Can we use multiple cables to validate the test method, so long as each one's specification is greater than 17lbs?

Any ideas on how to to guarantee homogeneity of our subgroups?

Thank you everyone in advance.

Elsmar Forum Sponsor
M

#### mwb0585

Also, not too sure of how important it is, but the Test (cable tensile test) is used on incoming lots to determine if they meet the spec.

N

#### NumberCruncher

Hi mwb0585

I wish I could give you the definitive answer, but I have yet to find any definitive references, or indeed any references. All of the information that I can find refers to destructive testing of your own product.

In this case, if I were a cable manufacturer, I would be able to take samples at different times (the dies used to produce cable will slowly wear over time giving different thicknesses of wire), from different heat treatment batches, different batches of alloy and so on. As a manufacturer, I can get a measure of process spread.

However, as the customer, all I get is a single reel of cable, from a single production run. It will be as homogeneous as the production process is capable of making.

My personal opinion is that you won't get any significant variation from one part of the reel to the next. Cable is produced in very large quantities, and each roll will have been manufactured and heat treated in a single batch. I think the only variation you will find is where the cable fails due to a defect and you get a very low tensile strength result. These results should be deleted from the data set.

The only way I can see of getting sufficient variation is do what your colleague suggests and use different diameter cables. I would go with one cable that is slightly thinner and one that is slightly thicker than your usual choice.

My reasoning is that, I want to know that I can detect weak cable. If the cable is 'too strong', is that a problem? But if it's too weak...

I repeat, this is not some official guidance that I am quoting from, just my personal opinion. But I have been in the situation of trying to carry out non replicable MSA with a one sided specification. I was forced to artificially introduce variation by using the 'wrong' heat treatment for some parts.

NC

I

#### IHaveNoSpaceBar

Hello.

Everything NC said +1 on his response.

I would do my testing with different diameter cables. This would provide you the best results due to the part to part variation. If you use the same cable you won't get enough part to part variation. You may end up wasting a lot of time trying to get a GRR below 30%. As far as the limits go, I would just do a one sided with the minimum, which maybe 17 in your case. I have to agree with him with regards to tolerance, it should be 17 lb/f to infinity for limits.

N

#### NumberCruncher

Also, not too sure of how important it is, but the Test (cable tensile test) is used on incoming lots to determine if they meet the spec.

Hi mwb0585

I'm going senile. I have only just spotted this. You could try using the different batches of cable as different groups first. It would be simpler and cheaper than buying in extra cable.

Is that what you are suggesting in your original post? It looks from what you originally said, that you were simply taking a single roll of cable, cutting it into sections and calling each group a subgroup. " My idea was to take successive samples from the cable spool..." This implies taking samples from the same roll.

Having said that, I still doubt that you will see enough variation between rolls to get a decent R&R result, especially if you get all your cable from the same supplier. By all means, give it a go, but I wouldn't hold my breath.

NC

M

#### mwb0585

Thanks for the responses, even though they go against my gut instinct.

Anyone have additional input?

What I love about this site is knowing that while this topic has been up for a day, it will remain up for years!

M

#### mwb0585

Yea, I mentioned that in the OP. We do have multiple lots of the same cable lying around, so it is possible. Still the concern that, because of within-lot variation (remind you, we are assuming that a single lot/spool is homogeneous... even though we know it isn't) the between-lot variation may not be picked up, and minitab may view them as non-distinct subgroups.

M

#### mwb0585

@Numbercruncher

Yea, I mentioned that in the original post. We do have multiple lots of the same cable lying around, so it is possible. Still the concern that, because of within-lot variation (remind you, we are assuming that a single lot/spool is homogeneous... even though we know it isn't) the between-lot variation may not be picked up, and minitab may view them as non-distinct subgroups.

D

#### debyang

If we focus on the GRR value instead of the %, will it still be an issue?